Selective encryption is a technique to save computational complexity or enable new system functionality by encrypting only a portion of a compressed bitstream while achieving adequate security. Although suggested in a number of specific cases, selective encryption could be much more widely used in consumer electronic applications, ranging from mobile multimedia terminals to digital cameras, were it subjected to a more thorough security analysis. We describe selective encryption and develop a simple scalar quantizer example to demonstrate the power of the concept, list a number of potential consumer electronics applications, then describe an appropriate method for developing and analyzing selective encryption for particular compression algorithms. We summarize results from the application of this method to the MPEG-2 video compression algorithm.

We introduce a method for locally optimal variable-to-variable length source coding with distortion, and apply it to coding the linear predictive coefficients of speech. The method is similar to entropy-constrained vector quantization, but it uses a dynamic programming algorithm to encode. The method automatically discovers variable-length source structure, in this case the acoustic-phonetic structure of speech. Using this structure, it is possible to compress the linear predictive coefficients of speech to one-third the rate of entropy-constrained vector quantization of speech, with no increase in spectral distortion. Auditory tests reveal that using this method, the spectral component of speech can be coded naturally and intelligibly to as low as 50 bits per second.>

The performance of a vector quantizer can be improved by using a variable-rate code. Three variable-rate vector quantization systems are applied to speech, image, and video sources and compared to standard vector quantization and noiseless variable-rate coding approaches. The systems range from a simple and flexible tree-based vector quantizer to a high-performance, but complex, jointly optimized vector quantizer and noiseless code. The systems provide significant performance improvements for subband speech coding, predictive image coding, and motion-compensated video, but provide only marginal improvements for vector quantization of linear predictive coefficients in speech and direct vector quantization of images. Criteria are suggested for determining when variable-rate vector quantization may provide significant performance improvement over standard approaches.>

Vector quantization (VQ) followed by entropy coding is used to compress linear predictive coefficients (LPCs) of speech into a variable-rate representation with distortion. It is shown in the LPC case that when a conditional entropy coder is used i.e., when the entropy code used for the current codeword is conditioned on the previous codeword, then a conditional version of entropy constrained vector quantizer (ECVQ) outperforms a conditional version of the straightforward approach by over 27%. Thus, conditioning restores the usual gain of ECVQ over standard VQ. This 27% reduction in bit rate is over and above the 42% reduction in bit rate already obtained by using a conditional rather than memoryless entropy coder in the straightforward approach. The product of these reductions is an overall reduction of 60% in average rate from the original, fixed-rate LPC system.>

In a decentralized hypothesis testing network, several peripheral nodes observe an environment and communicate their observations to a central node for the final decision. The presence of capacity constraints introduces theoretical and practical problems. The following problem is addressed: given that the peripheral encoders that satisfy these constraints are scalar quantizers, how should they be designed in order that the central test to be performed on their output indices is most powerful? The scheme is called cooperative design-separate encoding since the quantizers process separate observations but have a common goal; they seek to maximize a system-wide performance measure. The Bhattacharyya distance of the joint index space as such a criterion is suggested, and a design algorithm to optimize arbitrarily many quantizers cyclically is proposed. A simplified version of the algorithm, namely an independent design-separate encoding scheme, where the correlation is either absent or neglected for the sake of simplicity, is outlined. Performances are compared through worked examples.>

The authors have experimented with a variety of analysis/synthesis systems in the presence of quantization. The systems were block discrete Fourier transform (DFT), block DFT with trapezoidal windowing, block discrete cosine transform, Hanning windowed short-time Fourier transform (STFT) oversampled by a factor of two, Kaiser windowed STFT oversampled by a factor of four, quadrature mirror filter (QMF) banks, perfect-reconstruction QMF banks, pseudo-QMF banks, and Princen-Bradley filter banks. All systems were evaluated for speech compression; a subset was evaluated for image compression. It is shown that for speech the more sophisticated systems (e.g., QMF) provide the best objective performance, but comparable subjective performance can be obtained at a fraction of the complexity using simpler systems (e.g. pseudo-QMF and Princen-Bradley filter banks). For images, a tree-structured QMF bank did not perform well for 16*16 frequency bands when compared to block-oriented analysis/synthesis systems.>

The authors present a psychophysically justified bit allocation algorithm for use with subband image coding systems. This algorithm is also appropriate for DCT and DFT (discrete cosine and discrete Fourier transform)-based analysis-synthesis systems, as well as many others. The subjective performance of the algorithm was compared with the performance of a related algorithm that seeks to minimize the mean-square coding error. A two-dimensional analysis-synthesis system of J.P. Princen and A.B. Bradley (IEEE Trans. on Acoust. Speech, and Signal Proc., vol.ASSP-34, no.5, p.1153-1161, Oct. 1986) was used for the comparison. It is found that the psychophysically justified algorithm is superior to the minimum-mean-square-error algorithm at low bit rates.>

An algorithm introduced by L. Breiman et al. (1984) in the context of classification and regression trees is reinterpreted and extended to cover a variety of applications in source coding and modeling in which trees are involved. These include variable-rate and minimum-entropy tree-structured vector quantization, minimum expected cost decision trees, variable-order Markov modeling, optimum bit allocation, and computer graphics and image processing using quadtrees. A concentration on the first of these and a detailed analysis of variable-rate tree-structured vector quantization are provided. It is found that variable-rate tree-structured vector quantization outperforms not only the fixed-rate variety but also full-search vector quantization. The successive approximation character of variable-rate tree-structured vector quantization permits it to degrade gracefully if the rate is reduced at the encoder. This has applications to the problem of buffer overflow.>

The authors consider how much performance advantage a fixed-dimensional vector quantizer can gain over a scalar quantizer. They collect several results from high-resolution or asymptotic (in rate) quantization theory and use them to identify source and system characteristics that contribute to the vector quantizer advantage. One well-known advantage is due to improvement in the space-filling properties of polytopes as the dimension increases. Others depend on the source's memory and marginal density shape. The advantages are used to gain insight into product, transform, lattice, predictive, pyramid, and universal quantizers. Although numerical prediction consistently overestimated gains in low rate (1 bit/sample) experiments, the theoretical insights may be useful even at these rates.>